The Home of RLEnvironments

from openai import OpenAI
from openreward import OpenReward
import json

or_client = OpenReward()
oai_client = OpenAI()
MODEL_NAME = "gpt-5.4"

environment = or_client.environments.get(name="kanishk/EndlessTerminals")
tasks = environment.list_tasks(split="train")
tools = environment.list_tools(format="openai")

example_task = tasks[0]
with environment.session(task=example_task) as session:
    rollout = or_client.rollout.create(
        run_name="EndlessTerminals-train-quickstart",
        rollout_name="example_task",
        environment="kanishk/EndlessTerminals",
        split="train",
        print_messages=True
    )

    prompt = session.get_prompt()
    input_list = [{"role": "user", "content": prompt[0].text}]
    rollout.log_openai_response(input_list[0])
    finished = False

    while not finished:
        response = oai_client.responses.create(
            model=MODEL_NAME,
            tools=tools,
            input=input_list
        )
        rollout.log_openai_response(response)

        input_list += response.output

        for item in response.output:
            if item.type == "function_call":
                tool_result = session.call_tool(item.name, json.loads(str(item.arguments)))

                reward = tool_result.reward
                finished = tool_result.finished

                tool_result_item = {
                    "type": "function_call_output",
                    "call_id": item.call_id,
                    "output": tool_result.blocks[0].text
                }
                input_list.append(tool_result_item)
                rollout.log_openai_response(tool_result_item, reward=reward, is_finished=finished)

                if tool_result.finished:
                    finished = True
                    break

    or_client.rollout.close()